From forge
Optional research skill for gathering external context โ library docs, best practices, patterns, and prior art. Uses WebSearch, WebFetch, Context7, and GitHub doc search. Can be invoked standalone or as part of brainstorming/planning.
npx claudepluginhub ekelhaft-tools/forge-cursorThis skill uses the workspace's default tool permissions.
Gather external context to inform brainstorming and planning decisions. This skill uses web search, documentation fetchers, and GitHub repo search to find relevant information about libraries, patterns, APIs, and best practices.
Creates isolated Git worktrees for feature branches with prioritized directory selection, gitignore safety checks, auto project setup for Node/Python/Rust/Go, and baseline verification.
Executes implementation plans in current session by dispatching fresh subagents per independent task, with two-stage reviews: spec compliance then code quality.
Dispatches parallel agents to independently tackle 2+ tasks like separate test failures or subsystems without shared state or dependencies.
Gather external context to inform brainstorming and planning decisions. This skill uses web search, documentation fetchers, and GitHub repo search to find relevant information about libraries, patterns, APIs, and best practices.
This skill is always optional. It's offered as a choice during brainstorming and planning, or invoked standalone via forge research.
$ARGUMENTS: If the first token is a number or short identifier, use it as the session ID.ls .forge/sessions/ 2>/dev/null | sort
1), then mkdir -p .forge/sessions/1/<id>"SESSION_DIR = .forge/sessions/<id>/ โ all file paths in this skill are relative to this directory.
| Tool | Use Case |
|---|---|
WebSearch | General queries โ best practices, comparisons, "how to X in Y" |
WebFetch | Read specific pages โ docs, blog posts, GitHub issues, Stack Overflow |
mcp__plugin_context7_context7__resolve-library-id โ query-docs | Library documentation โ API reference, usage examples, configuration |
mcp__zread__search_doc | GitHub repo docs, issues, and commits |
From the user's input (or from brainstorm/plan context), extract specific questions:
If the user's request is vague, ask using AskQuestion:
Run research queries in parallel where possible (multiple WebSearch calls, Context7 + WebSearch simultaneously).
Library docs flow:
resolve-library-id with the library namequery-docs with the specific topic/questionGeneral research flow:
WebSearch with targeted queryWebFetch on the best results for deeper contentGitHub repo flow:
search_doc on the relevant repo with the questionCombine results into a structured summary. Focus on:
Save to $SESSION_DIR/research.md:
# Research: <topic>
## Questions Investigated
- <question 1>
- <question 2>
## Findings
### <Finding title>
- **Source**: <url or "Context7: library-name">
- **Relevance**: <why this matters for our task>
- **Key takeaway**: <1-3 sentences>
- **Code example** (if applicable):
```typescript
// relevant snippet
Create `$SESSION_DIR` if it doesn't exist. If `$SESSION_DIR/research.md` already exists, ask whether to overwrite or append.
### 5. Report Back
Summarize findings concisely to the user. If called inline from brainstorming/planning, return focus to that skill's flow.
## Rules
- **Max 5 WebSearch queries** per research session โ be targeted, not exhaustive
- **Max 3 WebFetch calls** โ pick the most relevant sources
- **Don't copy content verbatim** โ summarize in your own words, reference sources
- **Prefer official docs** over blog posts over Stack Overflow
- **Note freshness** โ flag if docs might be outdated (check dates)
- **Stay focused** โ research the specific questions, don't go down rabbit holes